Nearly 20 years have passed since Ed Lewis revealed the importance of Hox genes in the specification of different segments in the anterior-posterior axis of the fly. Pioneering studies by several authors, among others Garcia-Bellido and his student Ginés Morata, helped to elaborate a theory of segmental specification that was strengthened with the arrival of molecular techniques to the field of Developmental Biology. The conservation of Hox genes in metazoans at the level of sequence, function and complex organization has resulted in the export of this Drosophila theory as a paradigm to interpret the development of axial specification in organisms less amenable to experimental study. There are two main ways to interpret how Hox genes work in Drosophila. One considering Hox genes as "segment identity" factors giving global properties to the segments in which they are active. Another considering Hox genes as encoding spatially restricted transcription factors required for a number decisions taken at the cellular level. Here I use published and unpublished experimental data to illustrate that early activation of the Hox genes does not establish a gene code that leads to "segment identity". I will stress the point that Hox expression patterns develop with the embryo, that there are many genes involved in this modulation, and that the changing pattern of expression is important to achieve the final shape of the animal. I will show that, by interpreting Hox gene function in this way, some apparently paradoxical results in the Hox field can be answered. Finally, I discuss the implications of dynamic Hox gene expression on the evolution of segment morphology.